Description
Date depot: 1 janvier 1900
Titre: CoModels, engineering dynamic compositions of coupled models to support the simulation of complex systems
Directeur de thèse:
Christophe CAMBIER (UMMISCO)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
There are no methodologies today that can support modelers in:
● Representing and reasoning about these “composite” models
● Implementing and experimenting these assemblages of models
Some technologies have however been proposed in the latest years, but they are mostly oriented
towards proposing technological solutions to the operational coupling of models (HLA, DEVS) [1][2],
leaving aside the “semantic” problem of their static or dynamic composition. In particular, there is no
way one can provide, let alone revise and reuse, a description of their composition, such as, for instance,
the spatial and temporal scales of the models involved, their transfer functions, how they are supposed
to be combined during experiments, etc.
Consequences are: the scattering of models in the same disciplines, the difficulty to compare, to reuse
and combine them, and also the difficulty to reuse existing schemes for composing models.
We propose in the PhD to tackle this problem by simultaneously addressing the questions
of the modeling and simulation (execution) of these co-models (as a contraction of « combination of
models »).
The central claim of the thesis will be to consider a « co-model » as a model, and more specifically an
agent-based model, in which the agents wrap one or several instances of the models to couple, with
their own life-cycle, operations, collaborations, conflict resolution mechanisms, etc. which will draw
from the numerous works already published on multi-agent systems. A co-model, in this perspective,
will be a model that captures and represents a particular collaboration between these « sub-models »,
which can be based on existing collaboration schemes between experts, or any other way of organizing
their contributions. Conversely, this work will allow to consider « regular » agent-based models as very
specific implementations of co-models, where agents only wrap models of individuals.
The ultimate goal of the candidate will be to provide both a language and an infrastructure for co-modeling, and it is expected that, while reusing results from domains ranging from software engineering
to distributed simulation, innovative proposals will be made that will extend beyond the field of
modeling.
The development of this infrastructure will be based upon the existing GAMA Platform [4], which
already provides modelers with the possibility to design complex, multiple scales, heterogeneous agent-based models [3]. It will benefit from its modular structure and will in turn enable it to easily use and
reuse legacy models
Doctorant.e: Huynh Quang Nghi